xiaochaoqu的专栏

最近一直在看卷积神经网络（CNN），把我对CNN的理解记录如下（主要是在computer vision 领域）。本质上， 卷积神经网络是一种特殊的神经网络（NN）。CNN 与 NN 最大的不同在于结构的不同。对于NN来说，层与层之间的神经元是完全连结的(fully connected),  这样会导致可调节参数过多，学习这些参数非常困难，容易出现过拟合的问题。另外一个问题这种完全连结的结构

MATLABfunction matrix = zigZag(n)  %This is very unintiutive. This algorithm parameterizes the %zig-zagging movement along the matrix indicies. The easiest way to see %what th

Linear mapping, linear transformation, homomorphism, isomorphism, operators, linear and nonlinear transformations, change of basis, similar matricesDef. Linear mapping (or linear transformation)

http://www.intechopen.com/books/face-recognition

ppt地址 https://docs.google.com/presentation/d/1I-F2F9ntZzfZTk15yBcpMHEMqBK3VNezVOmiU2kKGnA/edit#slide=id.i86

https://sites.google.com/site/mihailsirotenko/projects/convolutional-neural-network-class

http://www.cad.zju.edu.cn/home/dengcai/Data/ReproduceExp.html

http://sugiyama-www.cs.titech.ac.jp/~sugi/software/SoftwareThe software available below is free of charge for research and education purposes. However, you must obtain a license from

读书笔记 ill conditioned and well conditioned linear systems ()

http://blog.163.com/lipse_huang/blog/static/19165754520133954138888/http://blog.youkuaiyun.com/zouxy09/article/details/14222605http://www.codeproject.com/Articles/16650/Neural-Network-for-Recognition-

1 . 课程 ：ICS273A  machine learning 地址： http://www.ics.uci.edu/~dramanan/teaching/ics273a_winter08/  Regularization and PCA 部分很好2. 神经网络 backprop algorithm Donald R. Tveter, The Backprop Algo

http://courses.cs.tau.ac.il/Caffe_workshop/Bootcamp/pdf_lectures/

When Face Recognition Meets with Deep Learning: an Evaluation ofConvolutional Neural Networks for Face Recognition

1. visualizing higher-layer features of a deep network本文提出了两种可视化方法。 1. 最大化activation当训练完一个深层神经网络之后， 固定所有参数。 然后对于某一个神经元的activation进行梯度上升优化来寻找能使它的值最大化的input。 不断的用gradient ascent来更新一个初始化为random

http://web.mit.edu/emeyers/www/face_databases.html#ar

人脸识别数据集

The following is the list of some good sentences from CV papers, more sentences will be added in the future. 1. In practice, the feature dimension m is often very high and, thus, it is necessary and

x = [2.5 0.5 2.2 1.9 3.1 2.3 2 1 1.5 1.1]; % 原始的数据有两个维度y = [2.4 0.7 2.9 2.2 3.0 2.7 1.6 1.1 1.6 0.9]x = x - mean(x); % 先减去均值，使得数据变为均值为0y = y - mean(y);plot(x,y,'*')data = [x' y'];C = cov(d

https://files.nyu.edu/mrg217/public/mle_introduction1.pdf

Distance in standard unitsIn statistics, we sometimes measure "nearness" or "farness" in terms of the scale of the data. Often "scale" means "standard deviation." For univariate data, we say that

来源：水木社区人工智能版。发信人: Insomnia (完美主义是种病), 信区: AI标  题: Machine Learning书单发信站: 水木社区 (Fri Mar 29 16:46:37 2013), 站内持续更新，请补充。除了以下推荐的书以外，出版在Foundations and Trends in Machine Learning上面的survey文章

These three types of matrices often form the basis of a multivariate method. The correlation and the covariance matrix are also often used for a first inspection of relationships among the variables o

地址  http://www.snl.salk.edu/~shlens/pca.pdf

The example of this section has three important properties.• Its variables (the amounts of ice cream and butter to produce) are continuous variables. They can take on any real value, subject to sati

http://www.imagemagick.org/discourse-server/viewtopic.php?t=22999&p=96286

a zero mean random vector is said to be white if its elements are uncorrelated and have unit variance

matlab 人脸识别代码

使用 \xleftarrow or \xrightarrow. 两者的用法相同。a=b \xleftarrow[H]{\xi+a\timesc} f=g          使用 \overseta=b \overset{F}{\longleftarrow}c=d          使用 \stackrela=b \stackrel{F}{\longleftright

matlab tool box for LDA-L1 norm

编译MatConvNet的时候，matlab 提示不能链接gdiplus库解决方法： 下载gdipllus库，并修改visual studio项目属性使得该库可以被链接。